The proposed research seeks to use meta-analytic methods to exploit the rather considerable databases on coronary artery bypass graft surgery, lytic agents for myocardial infarction, carotid endarterectomy, and drug interventions for congestive heart myocardial infarction, carotid endarterectomy, and drug interventions for congestive heart failure created during the first two years of NIH support. We are particularly interested in the clinical and methodological issues related to the generalizability or "external validity" (Cook & Campbell, 1979) of the results of meta-analysis. The basic question addressed by external validity concerns the scope and limitations of an observed effect. Essentially, this involves demonstrating that the effects are consistent for subgroups or clusters resulting from variations of treatments within different time periods, populations, or settings. Fortunately, the quantitative methods provide tests for heterogeneity. When such tests fail (i.e., indicate statistically significant heterogeneity), these external validity issues should be considered in accounting for this variation (although in some cases, heterogeneity may reflect methodological artifacts). If significant non-artifactual variation is detected, then an explanation for the differences must be determined. There are three analysis strategies. The first two that we have followed involve forming subgroups of individual studies or subgroups within studies based on relevant variables and determining if these strata account for the observed heterogeneity. This is consistent with the traditional approach to meta-analysis as developed by Glass where the categories or strata were established a priori. Another methodological approach involves the use of multiple regression to relate background variables such as health status to outcomes (here effect sizes). We have taken two approaches to examining treatment variation. First (as noted in Section C) we have expanded our database to allow inclusion of a variety of new treatments. Second we will conduct a temporal synthesis to examine trends (by cumulating results over time). Such a strategy not only allows us to track changes between similar treatments but also to note major changes within a technology (as with t- PA in managing acute myocardial infarction). Our preliminary research (reported in Section C) clearly demonstrates that the above approaches can be easily accommodated using our current databases and should be fully investigated before resorting to other procedures.